On the origins of entanglement constraints - Macromolecules (ACS

Comment on “Chain Entanglements in Polyethylene Melts. Why Is ... A Rheological Investigation of Entanglement in Side-Chain Liquid-Crystalline Azobe...
0 downloads 0 Views 1MB Size
Macromolecules 1993,26, 795-804

796

On the Origins of Entanglement Constraints D. Richter' Znstitut fiir Festkarperforschung, Forschungszentrum Jiilich, 51 70 Jiilich, Germany

B. Farago Znstitut Laue-Langevin, 38042 Grenoble, France

R. Butera3 L. J. Fetters, and J. 5. Huang Corporate Research Laboratories, Exxon Research and Engineering Company, Annandale, New Jersey 08801

B. Ewen Max-Planck-Institut fiir Polymerforschung, 6500 Mainz, Germany Received September 24, 1992; Revised Manuscript Received November 10, 1992

ABSTRACT We have studied the dynamic structure factors of poly(ethylenepropy1ene) alternating copolymer (PEP) and saturated polybutadiene (PEB-2). To investigate different models of entanglement formation, we varied the polymer volume fraction (PEB-2) and temperature (PEB-2 and PEP) and, thereby, systematically changed the two length scales considered to be important for the buildup of entanglements: namely,the contour length density and the random walk step length. In the framework of the scaling ansatz, our result is in the range of the binary contact models and excludes the packing model. Quantitatively, however,the scaling ansatz is not confirmed. Also,qualitative agreement is found with respect tothe topological calculations by Iwata and Edwards. Our results are at variance with one of the basic relations of the DoiEdwards theory of viscoelasticity, which relates the elastic response of a polymer melt with molecular chain parameters and the entanglement distance.

1. Introduction In the plateau regime of the dynamic shear modulus a dense polymer system responds elastically like a rubber. This behavior has been ascribed to the existence of a temporary network built up by the mutually interpenetrating long-chain molecules. The role of the cross-links, thereby, falls to long-lived topological constraints or entanglementa.1J As for arubber, the elasticitythen arises from the entropy elasticity of the strands between the entanglementa. The molecular origin of entanglements is not well understood, but current thinking postulates them to originate mainly from the topological nature of longchain molecules as being flexible, nearly one-dimensional uncro888ble objects. The occurrence of entanglements then ia governedby two length parameters, the step length of the Gaussian random walk of the chains and the lateral dietancebetween chains determining the amount of chain contour length per volume.3-9 In the reptation mode1,lO~llup to now the most complete theory of the viscoelasticityof dense polymer systems,the topologicalconetraintaare modeled by a tube surrounding the coarse-grained chain profile. Polymer motions are restricted to the interior of the tube which after a terminal time is left by creep motion through its ends. The tube diameter d may be identified with the distance between entanglements. The Doi-Edwards theory of viscoelasticity," which ie baeed on the reptation model, relates the plateau modulus, a property accessible to rheological To whom correapondence should be addressed. Prabnt ad&- du Pont Mar~~hall Laboratory,3600 GraysFerry Ave., Philadelphia, PA 19146. +

measurements, to the microscopic tube diameter of the reptation model.12

where (Re2)= nNC,k2 is the mean square end-bend distance, C, the characteristic ratio, Io the average mainchain bond length, n the average number of bonds per monomer, M the molecular weight, N the degree of polymerization, and p the polymer density. Equation 1 is one of the fundamentalrelationshipsof the Doi-Edwards theory relating macroscopic viscoelastic properties to the microscopic chain confinement. Until recently, the microscopic picture of a network of entanglements or a tube confining the polymer motion has been mainly inferred from rheometry,l revealingthe viscosity and the plateau modulus GN', and from the anomalous diffusion behavior of long-chain molecules.l3 In 1990,applyingneutron spin-echo (NSE)spectroscopy, Richter et al." reported a fmt direct micrwopic observation of the entanglement distance d, which in NSE experiments reveals itaelf as an intermediate dynamic length beyond which the relaxationof density fluctuations is strongly impeded. In a recent paper15 we gave a detailed accoullt of these NSE experimenta and their interpretation. Here we present a study aimed at the origin of entanglement constraints. We report NSE experimenta on the temperature dependence of the dynamic structure h&rs of saturated polybutadiene (PEB-2) and poly(ethy1enepropylene) alternating copolymer (PEP). For PEB-2 we

0024-9291/93/2226-Q796$04.oo/o Q 1993 American Chemical Society

796 Richter et al. also varied the polymer volume fraction. The experiments

were supplementedby temperatwedependent viscoelastic and small-angleneutron scattering (SANS)investigations of PEP aimed at the plateau modulus and the Kuhn length IK = loC, (10 = bond length). With this experimental approach we systematicallychanged the parameters which are supposed to determine the entanglement formation. In varying the polymer volume fraction, we studied the dependenceof d on the contourlength density; in changing the temperature and thereby the Kuhn length, we investigated the effect of the random walk step length 1 ~ . In section 2 we outline different ideas on the molecular origin of entanglements. Two classes of models are discussed: (i) the scaling models try to understand the important features of entanglement formation in terms of dimensional analysis and scaling ideas. Since two independent length scalesappear to be involved,a largevariety of scaling models are possible. We present a generalscaling model,3 the packing model,4 and two binary contact models.+7 The second class envelops topological approaches where in terms of mathematical models entanglement constraints are calc~lated.~-~ Section 3 describes the experimentalprocedures,includingsamplepreparation and characterization,neutron spin-echo experiments,and rheological experiments. Section 4 presents our results on the two polymers, PEB-2 and PEP. NSE data on polymer melts for different temperatures (PEP, PEB-2) and in concentrated solution for different polymer volume fractions (PEB-2) are displayed. Furthermore, temperature-dependent viscoelastic data on PEP are shown. In section 5 we discuss the experimental results. The different models for entanglement formation are scrutinized in the light of the experimental temperature and density dependence of the entanglement distance. Relating the temperature dependence of the plateau modulus, the microscopic entanglement distance, the chain flexibility, and the density, we investigatethe relation between plateau modulus and tube diameter as predicted by the reptation model of viscoelasticity.ll Section 6 summarizes our work. 2. Models for Entanglement Formation

The interpretation of viscoelastic properties of dense long-chain polymers is baaed on the existence of an intermediate dynamic length scale-the entanglement length-which in analqgy to rubber-elasticity acta like the mesh of a temporary network. In the reptation theory11J2 this leagth becomes the tube diameter, signifying the lateral constrainta for polymer motion. Though the entanglement distance is of crucial importance of our understandingof viscoelasticityin polymers, its molecular origin is still not very well known. Different ideas have been brought forward in order to draw a microscopic or molecular picture of the entanglement phenomenon: Baeed on the concept of a dominating influence of chain contour length density, Grawley and Edwards proposed ageneral scaling anaab;3de Gennesand others6Jpromoted the idea of the amount of interchain contacts being the essential feature; and hnca,'6 Lin,I7and Kavaeealie and Noolandi' proposed a packing criterion for entanglement formation. Partially related to the interchain contact approach but conceptually different is the topological model by Edwards8 and Iwata and Edwatdet who treat the entanglement problem in terms of integral invariants. In the following we briefly outline the different concepts and work out the epecific predictions on the dependence of the entanglement dietance d on polymer density and segment length or characteristic ratio C,.

Macromolecules, Vol. 26, No. 4, 1993 2.1. The Scaling Model. Graessley and Edwards3 assumed that large-wale interactions in dense polymer systems should relate only to the chainlike structure of the molecules, the essential effect being the topological interaction arising from the mutual uncrossability of chains. The important quantity, therefore, is the chain contour length per unit volume (LlV). In polymer networksthe modulus GlkTreflects the cross-linkdensity. In terms of the rubber analogue the plateau modulus GN' should relate to the interaction density and be largely determined by the contour length density. More bulky chainslike polystyrene (PSI correspond to a lower contour length density exhibiting a smaller GN', while slim chains like polyethylene (PE)have a high contour length density and consequently a high plateau modulus. To construct scaling relations, dimensionless quantities are required. They may be obtained in considering the length characteristic for the polymer conformation, the Kuhn length IK = CJo. With the contour length density LIV = YL,where v = N&dM is the number of chainslvolume (Na = Avogadro's number, 4 = monomer concentration, p = density) and L = Mlolmo (mo = molecular weightlbond), the scaling relation between the plateau modulus and the contour length density then reads

-

-

Considering further Y L ~ K ~4 and the experimental relation GN' 4a(witha between 2 and 2.3), eq 2 assumes the form of a power law (3) using the relation of Y, I K , and L to molecular quantities (see above) they arrive at (4)

Finally, with eq 1 for d we obtain (5)

2.2. Packing Models. Packing models relate the occurrence of an entanglement to the gradual buildup of geometricalhindrancesdue to the presenceof other chains. More precisely,entanglementsare determinedby a volume through which a certain number of other chains have to pass, or a mean number of neighboring chain segments belonging to other noninterrupted chains (chain ends do not count) is required to restrict the lateral degree of freedom. Thia approach is baaed on the obeervation that for many polymers the product of the density of entanglement strands n, = pNJM,, where Me = Nono is the entanglement molecular weight, and the volume spanned by the entanglement distance is roughly c0wtant16J7

K a v d s and Noolandi*have extended the theory and take explicitly into account the effectof chain ends, which adds a factor (1- MJM) to eq 6. The modifid theory describes the transition from entangled to nonentangled behavior at Me in terms of amean field second-orderphase transition. In replacing p by ( p 9 ) in eq 6, the theory may also be extended to concentrated solutions. Neglecting

Macromolecules, Val. 26, No.4, 1993

On the Origins of Entanglement Constrainta 797

the influence of chain ends, eq 6 yieldsfor the entanglement distance (d2 = Cmlo2Ne) (7)

We note that as a consequence of the packing criterion, the entanglement distance increases with increasing tendency of a chain to coil-coiling diminishesthe presence of other coils and the entanglement volume increases. Furthermore, as a result of packing, the length of an entanglement strand Ne d2 is inversely proportional to the bond density. Packing is a special case of general scaling for a = 3. 2.3. Binary Contact Model. A longstanding alternative scaling approach designed to grasp the nature of an entanglement is that, where an entanglement is defined by a certain fixed number of binary contacts along the chain.6~' This argument again is of a topological nature dwelling on the noncrossability of polymer chains. Let cp = N&/(mCm) be the density of Kuhn segments; then the number of Rb of binary contacts/volume is given by

-

-

nb (p21,3 (8) For the number entanglement strands n,/volume we have

The desired number of binary contacts per entanglement strands is then

Rb = const Re

- (&

mo)ltCmNe

Using eq 1 for the entanglement distance

-

1 (p4/m,)1, is obtained (general scaling a = 2). Recently,Colby and Rubinstein'* proposed an alteration of this scaling m t z , conjecturingthat an entanglement was determined by a constant number of binary contacts in the entanglement volume d3, bringing thus together packing and contact models. The scaling relation for the entanglement distance then reads (a = 1 d2 Cm2/31,2(p4/m0)4/3 We further would like to note that the apparent inability of simple scaling to give a definite answer relates to the fact that the entanglement problem considered as a geometricalphenomenon containstwo independent length scales-the step length of the random walk IK and the interchain distances given by the contour length density (Ll V)-lI2. Therefore, besides scaling arguments, further aseumptions as explained above are necessary. 2.4. The Topological Approach. Like the scaling approach, the toplogical calculations are based on the assumptionthat the entanglement problem may be purely geometrical,i.e., a mathematical problem, where dynamic effects do not play a role. The topological calculationsgo beyond scaling in that the topological invariants of the geometrical constraints are calculated rather than conjectured as in the scaling arguments. As entanglements are conaidered to be a purely geometrical effect, the entanglement distance can only be a function of the two length d e s in the problem-the step length of the random walk of the chains 1~ and the distance between chain d2

-

contours (L/V)-1/2. Using the Gaussian topological invariant, which counts the windings swept out by one curve around another, Edwards* calculated the dependence of the tube diameter on the chain contour density for two limiting cases. For pure random flight polymers 1 d--- 1 (13) lK(L/V) (&mO)lK is obtained. This result agrees with a scaling argument given earlier by Doi19 and corresponds to the GraessleyEdwards scaling formula for a = 3 which is also followed by the packing models. The other limiting case concernslocally smooth chains. For the case of a Gaussian chain in a network of rode, Edwards found d (L/V)-l/Z, which agrees with the abovediscussed binary contact model. Finally, considering wormlike chain bridges, the differences in the power laws between the two limiting cases and exponents between -1 and J/2 may be obtained. Recently, Iwata and Edwardss extended the Gaussian integral method to so-called localized Gaussian integrals. In this approach the chain is subdivided consistently into so-called local chains. Their interaction can be treated in terms of two-body invariants like the Gaussian integral while the multibody encounters of the full chains cannot be easily identified by this method. These calculations are the most advanced in relating chain properties to topological hindrance. The theory introduces a new quantity, the topological interaction parameter q, which measures the capability of a chain to entangle and may be considered as a property individual to each chain as the characteristic ratio. is mainly determined by the diameter of a polymer chain. As one would have expected, polymers entangle more if the chain diameter is small or the contour length density is high. Furthermore,f appears to be also related to the size of Cm-the larger C, becomes, the more a chain reaches out in space and the better it entangles. An analytic functional dependence, however, cannot be read off from their reeults. Finally, for the concentration dependence of the plateau modulus they where a increases with decreasing calculate GN concentration (1.97 Ia I2.2). For high concentrations their result agrees with d ( p $ / m 0 ) - 1 / 2 . Besides this mathematicalapproach,there are ale0more heuristic topological considerations under discussion, A typical model of this class is by W U , who ~ considered entanglements as knots or hooks between chains. Here, it is believed that with increasing chain flexibility the chains will be able to form hooks more easily. If Izh = l/(Cmk,) is the number of node pointa per length along the chain, then the probability of a hook between chaine is proportional to nh2 (two chains have to have a node at the same position). Therefore, the arc length between two entanglementsshould beLe2 l/nh2.With that we arrive at Ne Cm2or

-

-

-

-

-

d2 Cm3 (14) We consider this ansatz as inadequate since it neglects completely the effect of contour length density. Nevertheless, comparing data for different polymers, AharonP has deduced a correlation of the form Ne (2-2, a result contrary to Lin's analysis.

-

3. Experimental Section 3.1. Sample Preparation and Characterization. The polyisoprenes and polybutadienes were prepared via standard anionic polymerization techniques: tert-butyllithium, purified by sublimation, was the initiator. The deuterated monomers

Macromolecules, Vol. 26, No.4, 1993

798 Richter et al. Table I Polymer Molecular Characteristics sample M , x 10-4 a MZlMwb MwIMnb 8.38 1.03 1.05 PEP-dlo 8.22 1.02 1.05 PEP-hio 36.70 1.03 1.04 PEP-hi0 I11 7.05 1.04 1.04 PEP-2-he 7.32 1.02 1.04 PEB-2-ds 0 Light scattering. b Size exclusion chromatography. were obtained from Cambridge Isotopes. The microstructure (H- or D-NMR) of the polyisoprenes was 75% cis-l,4, -18% trans-l,4, and -7% 3,4 while that of the polybutadienes was -40% cis-1.4, -53% trans-l,4, and -7% 1,2. Size exclusion chromatography was used to evaluate the heterogeneity indices. The saturation by hydrogen or deuterium of the polydienes was done by using palladium on calcium carbonate. The procedure followed that of Rachapudy et al.22 NMR analysis showed that saturation levels of >99.7% were obtained. The saturation of polyisoprene yielded esaentidy alternating amorphous poly(ethylenepropy1ene) (PEP), while polybutadiene yielded the crystalline poly(ethy1ene-1-butene) copolymer which is designated as PEB-2 (where the integer denotes the approximate number of ethyl branches per 100 backbone carbons). Densities at 23.1 "C for the amorphous polymers were measured in a density gradient column. Weight-average molecular weights were evaluated via a Chromatix KMX-6 low-angle laser photometer. For the case of the PEB-2 materials, the molecular weights of the parent polybutadienes were measured and corrected for saturation. The sample characteristics are given in Table I. 3.2. Neutron Scattering. All neutron scattering experiments to be discusaed in the following were carried out with the aid of neutron spin-echo spectro~copy.~~ In such an experiment, the neutron polarization P is measured as a function of the applied guide field H for various scattering angles ( 8 ) . For a coherently scattering specimen, P(Qmdirectly gives the normalized scattering function S(Q,t)/S(Q,O),where Q = (4r/X) sin I9 is the momentum transfer, X the neutron wavelength, and t the time, which is proportional to H and X3. The scattering contrast generally arises from the differences in scattering length of the deuterated matrix or the deuterated solvent and a protonated polymer. Under these conditions the coherent scattering of the individual chain or its pair correlation function is measured. We used sealed Nb containers which were loaded under inert conditions to avoid oxygen which could degrade the polymers at high temperatures. The samplea were heated in a vacuum furnace up to 600 K with a temperature stability of AT = f l "C. The backgrounds from the matrices-nearly entirely inelastic scattering-and from the container were measured separately and subetractedusing the proper transmissionfactore. Resolution corrections were performed using the instrumental resolution function obtained from a glassy PS standard (90/10 d/h) at room temperature. 3.3. Bheological Measurements. Rheological measurements were performed on a high molecular weight PEP sample (PEP111)(see Table I) using a Rheometrics System 4 rheometer with 25-mm-diameter plates in the mcillatory shear mode, with gap spacings which ranged between 1 and 2 mm. The sample was surrounded by a nitrogen atmosphere during all measurements. Isothermal frequency scans were performed using frequencies between 10-3and 100 rad/s. The strain amplitude was kept as small as possible, but ranged between y = 0.02 and y = 0.5, depending upon the frequency and temperature being used. Linearity of response was checked whenever the strain was increased, and all data reported are within the linear regime. Adjustmente were made during each temperature change to account for thermal expansion of the plates and sample. The sample preparation and loading procedure were as follows: First the sample was dried to constant weight under vacuum at 75 "C,producing a smooth bubble-free sample. Approximately 1.1g of this material was compression molded between mylar sheets at 70 "C for 23 min. The mold was then cooled to room temperature under a slight pressure. The sample was released from the mold, placed on the lower plate, and squeezed between

the plates until a slight normal force was registered. The temperature was raised to 100 "C and held there for 20 min. While at 100 "C the sample was squeezed until it began to flow. It was then trimmed and raised to 150 "C for approximately 1 h to allow complete adhesion to the plates. The temperature was then decreased to 60 O C , the sample was trimmed, and the temperature was raised again to 150 "C;the sample was held at that temperature for 30 min before starting the measurements.

4. Results

To pursue the problem of entanglement formation, a seriesof NSE experimentswere performed on PEB-2where the polymer volume fraction and temperature were changed. These experiments were supplemented by a temperature-dependent study of PEP, where we also investigatedthe plateau modulus. PEB-2 (essentiallyPE with one ethyl branch every 50 main-chain bonds) combines a very fast Rouse relaxation with strong topological constraints15and is therefore very well suited for a NSE study, which is always limited by spectral resolution. For PEB-2 it was possible to access a large concentration and temperature range using neutrons of short wavelength (X = 8.5 A) covering a time window 0.3 It I17 ns. At this wavelengththe intensity is about 4 times higher than that at X = 11 A,which was used in earlier NSE studies on melt dynami~s.14J5~24.25 For the dilution experiments we used the oligomer CIgDa, which can be considered as athermal. The deuterated paraffii was obtainedfrom CambridgeIsotopes and was 98 % deuterated as verified by proton NMR. The diluted samples were produced in diluting a master alloy ofQO/lOd/hPEB-2withtheoligomer.Toattemptaproper background correction, we measured the scattering from 100% deuterated PEB-2 and a pure C19Da sample. The background turned out to be purely inelasticand increasing with temperature. The subtraction was performed in combining respective proportions of the scattering from these two samples. To estimate whether the paraffin causes any SANS when contrasted with the protonated PEB-2 fraction in the mixed samples, we estimated its scattering contribution from an RPA formula for a threecomponent mixture, neglecting any specific interaction effects.26 In the worst case at 75% dilution and at Q = 0.15 A-l, the SANS contribution from the paraffin scattering is estimated to be below 3 7% of the total signal, rendering it negligible. Experimentswere performed in the concentration range 0.25 5 4 I1 polymer volume fraction at T = 509 K. Temperature-dependent measurements were carried out at 4 = 1 and t#~ = 0.5 covering a temperature range 418 I T I556 K. At each concentration and temperature, spectra of at least three different Q values were recorded. Thereby, depending on temperature and concentration, a Q range 0.052 IQ I0.155 A-1 was covered. Figure 1 presents representative spectra at three different polymer volume fractions. To directly visualize the effect of entanglement constraints the data are plotted against the dimensionless scaling variable u = Q2P(Wt)llZ of the Rouse m ~ d e l , ~ where ~ ~ B1 is the segment length, W = 3kT/(t12)(t = segmental friction coefficient), and t is the time. In such a presentation the dynamic structure factor originating from Rouse motion collapsesto a master c ~ r v e . ~ 5 The 3 , ~ Q-dependent splitting arises from the existence of the entanglement length at an intermediate dynamic length scale and thus demonstrates the presence of entanglement c o n ~ t r a i n t s . ~ ~ J ~ The solid lines present the effective medium model by Ronca.16 This model describes the dynamic structure factor in the crossover regime from unrestricted Rouse

Macromolecules, Vol. 26, No.4, 1993

oassA-l]

0.21

- 0.8 O.

*

\

-

On the Origins of Entanglement Constrainta 799

s v) 0.6q 0.4-

2 0.2:

y

02t

0 0‘ - 2

I‘m

Figure 1. NSE spectra obtained from PEB-2 at 509 K for three different volume fractions in a Rouse scaling presentation: upper part 4 = 1; middle part 4 = 0.5; lower part 4 = 0.3. The solid lines are the results of fits with the Ronca model.16 motion to entanglement-controlledbehavior and contains two parameters: the tube diameter or entanglement distanced and the “Rouserate” WZ4. Thereby,the plateaus developingat large u signifythe amount of constraints-a high plateau value translates into a small entanglement distance and vice versa. Comparing spectra at similar Q values and different polymer concentrations, we observe a dradtic decrease of the constraints: e.g., at 4 = 1.0 the plateau for Q = 0.116 A-1 extra olates to about 0.6; at 4 = 0.5 the plateau for Q = 0.103 reaches already below 0.4, while at 4 = 0.3 this level has dropped to less than 0.1. We also note that even at a polymer volume fraction of only 0.3 we still find pronounced deviations from scaling with the Rouse variable. Thus even at this concentration a well-defined intermediate dynamic length scale arising from the entanglement constraints can be observed. Ail has been discussed earlierI5 and may also be seen from the upper set of spectra in Figure 1 for the PEB-2 melt, the Ronca model and the experimental results show somedisagreement. Similardeviationsoccur for the other aampleswithhigh polymer volume fraction (4 2 0.7),while they are absent for the two lower concentrationsdisplayed. In the earlier studyl5 we have also shown that applying the local reptation model of de Gennes,30 which considers the equilibration of density fluctuations along the fiied tube, a better agreement between experiment and theory may be achieved. The local reptation model gives an explicit expression for S(Q,t) to first order in (Qd)2

1-l

S(Q,t)/S(Q,O)= 1-

%+ %

exp($)

8

6

10

12

14

a? L zm Figure.2. Initial decrease of the dynamic structure factors from PEB-2 at volume fractions 4 = 1 , 4 = 0.5, and 4 = 0.3 in a Rouse scaling presentation. The solid lines representthe Rouse master function. The data points from the upper curve represent the Q values: 0.078, 0.116, and 0.155 A+; from the middle curve: 0.052,0.078, 0.116, and 0.155 A-l; from the lower curve: 0.052, 0.065, 0.078, and 0.116 A-l.

4 0 12 16 20 24 2% Q2

4

o2

erfc

(i)(15)

thereby, u = QV(Wt)1/2 is the Rouse variable defined above. Since the local reptation model neglects the shorttime unrestricted Rouse motion, this must be excluded in the analysis with eq 15. Consequently, the value of the Rouse rate WZ4entering into u = Q212(Wt)1/2isdetermined from the shorbtime part of the spectra. Figure 2 presents Rouse scaling plots of the initial part of the spectra for the same concentrations as shown in Figure 1. The solid line depicta the Rouse dynamic structure factor serving as a master curve. From fits of these data involving always only the Rouse rates as a fit parameter, the related monomeric friction coefficients f were obtained. For the

Table I1 Friction Coefficients and Entanglement Distances in Concentrated PEB-2Solutions and Melts at T = 609 K POlF vol

d

5

w14

(A)

frac cp (10-13A 4 s-l) [10-’0 (dyn cm) 8-11 de Gennes 1.00 7.0 0.7 4.0 0.4 43.5 0.7 0.90 9.2 0.8 48.0 0.5 3.0 0.3 1.9 0.2 55.1 0.4 0.70 14.3 1.5 0.60 11.7 0.8 2.3 0.2 59.9 0.7 1.5 0.1 0.50 18.2 1.1 67.0 0.6 0.92 f 0.03 0.35 29.8 1.1 83.6 0.3 0.30 33.7 1.2 0.82 f 0.03 102.6 1.8 0.74 f 0.03 -.0.25 37.2 1.5

** ** ** *

** ** *

l.OT I

I

** **

O

I

I

Ronca 42.8 0.9 45.5 0.7 52.8 0.6 58.9 1.3 62.0 0.7 77.5 1.4 88.4 1.5

-

* * * -

I

WllO

2

-2 1.0 v)

08

0.6 0.4

0.2

OO

10

20

30

40

a2 L2 IWf Figure 3. Rouse scaling representationof the long-timetails of the PEB-2 spectra at 4 = 0.9 and 4 = 0.5. The solid lines are fits with the local reptation model.30 determination of t from W14 = 3kTl2Itwe assumed l2 = C-102 = 13.06 A2 (Table 11). Keeping these Rouse rates fixed, the long-time taila ( t 27 of the spectra were then fitted with the dynamic structure factor of localreptation (eq 15). Figure 3displays the results for 4 = 0.9 and 4 = 0.5. Apparently, the local reptation model fits the spectra significantly better than the Ronca model. In particular, it accounta for the continuous, gradual decay of S(Q,t)/S(Q,O)even at large u. Because of the better accuracy of the de Gennes local

Macromolecules, Vol. 26, No.4, 1993

800 Richter et al.

/

80

z 7oc 1

a5 0.2

. I

1

1

100

?2

*

I

1

1

I

50

35

25

@ ["/.I Figure 4. Double logarithmicpresentationof the entanglement distance or tube diameter d for PEB-2 at 509 K as a function of polymer volume fraction 4: (0,solid line) results from a fit with local reptation; ( 0 ,dashed line) fit with the Ronca model.

reptation model,all entanglementdistances were evaluated in terms of this model-the values ford from both models differ only slightly, the maximum discrepancy occurring at the lowest polymer concentration. We would like to note that at low polymer volume fraction the crossover from unrestricted Rouse motion to local reptation shifts toward the outer part of the experimental time window, adding mme uncertainty to the evaluated results. Therefore we a h fitted all data sets in terms of the Ronca model, where all data points were considered. Figure 4 displaysthe results ford as a function of 4 (see also Table 11). The open symbols represent the outcome of the fit with eq 15 while the full circles are the Ronca results. The entanglement distances from the Ronca fit come out slightly smaller-on the average by about 5%-than those obtained from local reptation. Within experimental accuracy, however, the dependence on 4 is 4-0.61*o.02corresponds to identical. The solid line d local reptation while the dashed line d 4-O~ss*o.02 represents the Ronca results. At 4 = 0.3 in the case of local reptation the d value is definitely larger than suggested by the power law (solid line). The results from the Ronca treatment point in the same direction though the deviation is not as large. For 4 = 0.25 the Q-dependent splitting in the scaling representation of the data points is absent, and consequently an entanglement distance cannot be determined anymore with certainty. The observed deviation from the power law behavior at low 4 may result from the crossover to the semidilute regime, where a larger exponent (-3/4) is expected. This is further underlined by our failure to observe clear indications of constraints at 4 = 0.25, implying that there the tube diameter must be considerably larger than what would be extrapolated from the high-concentration range. We also addressed the temperature dependence of the microscopic dynamics of PEB-2. Figure 5 presents NSE spectra taken from the melt (4 = 1) at 418 and 483 K. Comparing the spectra qualitatively, we observe a strong increase in initial relaxation-the spectra at 483 K drop much faster than those at 418 K while the values reached at long timesdiffer only slightly. Obviouslythe topological constraints change only slightly. Thia is reflected in the experimental result far the temperature-dependent entanglement distance displayed in Figure 6 (eeealso Tables I11 and IV). Again the data points were obtdned by the two-step evaluationprooedure involvingshort-time Rouse and long-time de Gennes model fits. (For comparison in Table I11 we ale0 display the corresponding results from a fit with the Ronca model. Within experimental error they agree with the result of the local reptation model.)

-

i 1

T=418K

1

/"

5 oI t 40'

,

1 "'

-

T=483K 0'

4

8 12 Time [nsl

16

20

Figure 5. NSE spectra from PEB-2 for two different temperatures. The solid lines are results of a fit with the Ronca model: (A)Q = 0.077 A-*; ( 0 )Q = 0.116 A-l; ( 0 )Q = 0.155 A-l.

The results obtained for 4 = 1.0 and 4 = 0.5 are displayed in Figure 6. They are compatible with an identical temperature coefficient for both concentrations. The broken line represents the best fit for the 4 = 1 data d = 23.5 exp((l.2 f 0.2) X (16) Finally, the temperature-dependent Rwee rates or the monomericfriction coefficientsrespectivelywere evaluated inte~ofWLFshiftfactorslogcrT=co(T-To)/(T-T-). Within experimental accuracy, the WLF relation baeed (T,= 124 K, on the viscosity results from Carella et co = 6.35 for TO= 270 K)for their sample with the lowest methyl-branch content describes the temperature dependence of the microscopic Rouse rates. We also investigated the temperature dependence of the dynamic structure factor from PEP. Using a wavelength of X = 11.25 A, which offers an experimental time window of 0.7 It I36 ns, we recorded spectra at eight different temperatures over a temperature range 373 IT I598 K. At each temperature spectra at three different momentum transfers were measured (Q = 0.068,0.097, 0.126 A-9. The spectra were fitted in terms of the Ronca model, which provided a good description of the experimental spectra.33 Figure 7 displaysselected spectrataken at Q = 0.126 A-1 at different temperatures. The solid linea are the results of the respective Ronca fits which at each temperature were performed under the consideration of all three available Q values. The results are given in Table V. Again the spectra are characterized by two subsequent time regimes: an initial decay succeeded by a plateau region which becomes increasingly better separated with rising temperature. Higher temperature causes a strong acceleration of the initial relaxation and a decrease in the number of constraints-the plateau level appeara to drop with increasing T. Only for the spectra at the two highest temperaturea (548and 698 K), no further change in the long-time plateau level occurs. Figure 8 displays the resulting temperature-dependent entanglement distance. This parameter increases with T (d In d/dT = 2.8 X 10-3 K-1) until leveling off above 600 K.We note that the temperature coefficientof d for PEP L much larger than that for PEB-2 (Kpw = 2 . 3 K p ~ ~ Figure ). 8 ale0 includes results on the temperature dependence of the mean square radius of gyration (R;) C,(T) from PEP obtained by SANS." R,2 decreases with increasing temperature. Thereby, the temperature coefficient de-

-

Macromolecules, Vol. 26, No.4, 1993

On the Origins of Entanglement Constrainta 801

Table I11 Enknglememt Mstracer, h u m Rater, Characteristic Ratior, Dsnsitier, and Invariant of the Bubber Analogue for PEBd d (A)

T (K)

de Gennea

WI4 (10-13 A4 8-9 1.47 f 0.2 2.81 f 0.4 3.60 0.5 3.95 f 0.4 7.00 f 0.7 7.30 f 0.7 8.00 f 0.5

418 446 463 484

*

509 529 556

Ronca 38.4 f 1.2 39.9 f 0.7 42.6 f 1.0 40.2 f 0.7 42.8 f 0.9 43.5 f 0.6 44.2 f 0.7

38.5 f 0.6 39.8 f 0.5 42.2 f 0.6 40.9 f 0.6 43.5 f 0.7 44.1 f 0.4 45.8 f 0.7

C,”

6.52 6.32 6.19 6.05 5.87 5.74 5.57

P

(glcm3) 0.783 0.767 0.757 0.746 0.733 0.722 0.708

d2GNolpTC, 1.53 f 0.05 1.61 f 0.05 1.81 f 0.06 1.68 f 0.06 1.97 f 0.07 1.95 f 0.04 2.10 f 0.07

0 C, is based on the SANS results at 413 K for polyethylene and PEB-2. See ref 31; Lieae, G.; Fischer, E. W.; Ibel, K. J. Polym. Sci., Polym. Lett.Ed. 1975,13,39;andHortenlJ.C.;Squires,G.L.;Boothroyd,A.T.;Fe~rs,L.J.;Rennie,A.R.;Glinka,C. J.;Robineon,R.A.Macromolecules 1989,22,681. The combined data yields Rc = 0.45W.W,a value in excellent agreement with that from dilute solution measurementa under 8 conditions.

Table IV Entanglement Distances and h u o e Relaxation Rates for a 4 = 0.6 Concentrated Solution of PEB-2 in C@40 ~

~~~~~

T (K) 463

11.0 f 1.3 18.2 f 1.1 25.5 f 1.9

509 556 60

d (A) 65.1 f 0.8 67.0 f 0.8 72.7 f 1.1

w14(10-13A4 8-1)

C

l

-50

600

500

400

T [KI Figure 6. Temperature dependence of the entanglement distance d of PEB-2: ( 0 )c$ = 1; (0)4 = 0.5. The dashed lines are guides for the eye. The aolid line gives the prediction of the general ecaling model of Graessley and Edwardss (see text). I

I

1

8

1D 0.8

=

d, 0.6

-2 0.4 -

%? v)

o.2

t

I

I

well with the shift factor obtained from viscosity. This disagrees somewhat with our previous results.24 The difference originates from the explicit consideration of a temperature-dependent segment length, which was neglected in ref 24. To relate the measured microscopic entanglement distances to the viscoelasticbehavior of PEP, the plateau moduli of the large M, PEP I11sample were measured as a function of temperature. Data were obtained over the temperature range 298 IT I 548 K. After every scan at temperatures higher than 448 K,the sample was returned to 448 K and a frequency scan was performed at that temperature to examine possible thermal degradation. This was not found for any specimen and was confirmed by SEC. Attempts were made to acquire data at 573 K, but the sample was found to undergo slight thermal degradationas indicated by the subsequentfrequencyscan at 448 K (decrease in 70 and Gm”)and SEC analysis. Because of the narrow molecular weight distribution of the samplesused, G“possesaes a maximum Gm”,occurring at a frequency 0,. Gm” can be related to the plateau modulus GN’ according to ref 36, G,” = 3.56G~’.The G,” values were obtained by an interpolation procedure: the data near Gm” (usually 8-10 data points) were fit by a fifth-order polynomial,and the resulting polynomial was used to determine the maximum in the G” curve. Figure 10 shows the temperature dependence of Gm” for two differentPEP 111samples. The samplerepresented by the solid points was run in an effort to define more completely the high-temperature region, which shows an apparent plateau above 200-225 ‘C. The Gm”(7‘)values for each sample were normalized using the respective values of Gm” obtained at 175 “Cfor each sample. Table VI summarizes the values used to generate Figure 10.The small differences in absolute magnitude of Gm” (175) between the two samples (